The result of Komodo 5.1r1 is now online
http://ls-ratinglist.beepworld.de/
Stefan
(perhaps you have to clear your browsercache or reload the website)
LS-ratinglist: Komodo 5.1r1
Moderators: hgm, Rebel, chrisw
-
- Posts: 2460
- Joined: Sat Sep 03, 2011 7:25 am
- Location: Berlin, Germany
- Full name: Stefan Pohl
-
- Posts: 5966
- Joined: Sun Jan 10, 2010 6:15 am
- Location: Maryland USA
Re: LS-ratinglist: Komodo 5.1r1
Thanks. This does create a mystery for me. We found that due to reduced overhead the r1 version was MUCH stronger on 4 cores at time limits around game/1' (although this advantage pretty much evaporated at time limits like IPON uses or slower). But your result shows only 2 elo gain. So the question is why the huge disparity in results, since your time limit is close to game/1'?
Probably this means that the r1 version is much stronger at bullet speed on multi-core but not on single core. Also the interface may be a big factor in games played this fast. I used Fritz 11 interface; perhaps the overhead problems we fixed were not relevant on the interface you used.
Best regards,
Larry
Probably this means that the r1 version is much stronger at bullet speed on multi-core but not on single core. Also the interface may be a big factor in games played this fast. I used Fritz 11 interface; perhaps the overhead problems we fixed were not relevant on the interface you used.
Best regards,
Larry
-
- Posts: 2460
- Joined: Sat Sep 03, 2011 7:25 am
- Location: Berlin, Germany
- Full name: Stefan Pohl
Re: LS-ratinglist: Komodo 5.1r1
The LittleBlitzerGUI has no problems with output overhead, because it has nearly no output, while the games are runnig...lkaufman wrote:Thanks. This does create a mystery for me. We found that due to reduced overhead the r1 version was MUCH stronger on 4 cores at time limits around game/1' (although this advantage pretty much evaporated at time limits like IPON uses or slower). But your result shows only 2 elo gain. So the question is why the huge disparity in results, since your time limit is close to game/1'?
Probably this means that the r1 version is much stronger at bullet speed on multi-core but not on single core. Also the interface may be a big factor in games played this fast. I used Fritz 11 interface; perhaps the overhead problems we fixed were not relevant on the interface you used.
Best regards,
Larry
Stefan
-
- Posts: 5966
- Joined: Sun Jan 10, 2010 6:15 am
- Location: Maryland USA
Re: LS-ratinglist: Komodo 5.1r1
Thanks, that explains everything. So only those who used more popular interfaces at bullet or near-bullet speeds should retest the r1 version.pohl4711 wrote: The LittleBlitzerGUI has no problems with output overhead, because it has nearly no output, while the games are runnig...
Stefan
-
- Posts: 213
- Joined: Tue Jun 25, 2013 8:19 pm
Re: LS-ratinglist: Komodo 5.1r1
Over at Immortal chess Komodo 5.1r1 MP does not have good results at 5 min or faster in either Little Blitzer or Deep Fritz 13 vs Houdini 3 Pro